Harnessing AI in Education to Transform Student Experience

Whilst teaching and learning remains a human-led experience, AI can be a vital tool to help enhance these areas of education.

Disruptive technologies like generative AI (Gen AI) can be used to customise lesson plans for students based on individual needs. Likewise, it can be used to drive efficiency and streamline administrative tasks for educators.

One of the organisations harnessing AI to transform the education sector is Turnitin, a web-based plagiarism prevention tool that helps students and educators identify and correct plagiarism. In 2023, the organisation launched AI writing detection capabilities across its services to help educators identify when AI writing tools such as ChatGPT may have been used in writing submissions from students.

Leading Turnitin AI, Eric’s team is responsible for building intelligent systems that support learning outcomes. With more than 15 years of experience deploying enterprise-scale AI to drive transformation, he explains that AI helps “enable our educators and academic institution customers to uphold academic integrity for over 40 million students spanning more than 140 countries worldwide.”

In this exclusive interview with Technology Magazine, Turnitin’s Vice President of AI Eric Wang discusses the transformative impact of AI on Turnitin’s capabilities and the broader education sector.

How has your work with AI transformed Turnitin's capabilities? 

It’s safe to say that AI has been transformative to Turnitin, the entire education community and beyond, especially in the last few years. That said, teaching and learning are deeply human experiences and good AI in education should empower students and educators and promote moments of human connection and personal growth. 

To support those learning moments, we seek to give context to every decision in the learning process with the belief that better informed decisions lead to better student outcomes. From the widely used AI writing detection feature to the grading and feedback support tools, our AI models leverage the rich data and knowledge we’ve gathered over the last quarter century defining academic integrity.

Our AI writing detection capabilities helped educators and institutions find their footing when Gen AI massively and suddenly disrupted every aspect of education, reinforcing the vital role that Turnitin plays in global education and academic integrity. Looking forward, we are excited to share our workflow automation and proof of process products with the world - we think they will redefine how students create work with confidence and how educators can grade and give meaningful engagement and interaction between teachers and students that are indispensable to learning.

Tell us about Turnitin’s AI Writing Detector.

Our AI writing detection feature is designed to serve as a data point and provide information that educators can use to supplement their expertise when reviewing work submitted by students. It’s not an sole indicator that determines whether or not a student may have used Gen AI applications like ChatGPT, paraphrasing platforms or similar tools. Instead, it allows educators to start a conversation to determine the student’s critical thinking methods and ultimately help uphold academic integrity and originality. 

It’s crucial to recognise an educator’s experience can’t be substituted – they know their students and how they write best – technology should only ever support that.

We started our R&D work on detecting LLM writing back in 2020, just after GPT-3 launched. The critical research and learning around how LLMs write, especially how LLM writing differs from human writing, continues to inform how Turnitin AI writing detection works today. Our detection capabilities are actually multiple models that work together to identify not just the statistical signatures of LLMs, but also of AI paraphrase tools. They use the latest transformer architectures, which is the same architecture used by cutting edge frontier LLMs. 

Can you discuss some of the opportunities and challenges of developing this tool? 

AI is one of the most powerful and transformative technologies in our society. It’s rare to work on something that has this sort of impact. Turnitin AI takes a holistic approach to building AI into our products. AI can be a force of enormous good, as we think about the impact of our technology on the learning journeys undertaken by our students, educators and institutions, including the opportunity to provide individualised learning efficiently. 

One of the key challenges is building AI that amplifies the great things that are happening while also mitigating any harmful effects. When ChatGPT launched in November 2022, we knew Turnitin had an important responsibility and opportunity to help educators in this critical moment. It really was one of those once-in-a-career moments where you realise “If not us, then who?”

Our data further allows us to analyse and safeguard our model behaviours - for example all of our AI writing detection models undergo a battery of large scale pre-GPT human writing tests to get an accurate measure of false positive rates. We also have built a highly autonomous model training pipeline that creates complex and challenging training and testing datasets at scale, allowing us to innovate and iterate faster than ever before and keep pace with the rapidly changing AI landscape.

Of course, breaking new ground comes with its own set of challenges as well. Our top priority in building our AI writing detection feature was to ensure its safety. To do this, we set a very tight false positive rate target of below 1%, while still maintaining consistency and reliability. 

Another challenge was in helping to communicate our findings and the model’s behaviour and model performance to our users in a way that was both clear and actionable. We achieved both of these difficult goals through a highly collaborative cross-functional effort between our AI, Engineering and Product teams. The work was messy, complex and creative - it pushed all of us to excel in new ways and achieve new levels of impact. Turnitin emerged from this experience a stronger, faster and more capable team than ever before.

How does Turnitin address the ethical concerns of AI use in education? 

At Turnitin, our guidance is and has always been, that organisations make the final determinations on what is an acceptable use of AI usage. AI can serve as an operational efficiency or productivity tool and has incredible potential to benefit humans and society, but it must be developed thoughtfully. 

Discussions and policies around ethical and responsible AI aim to help people build socially beneficial, safe and accountable technology. Organisations should consider focusing on AI literacy training, putting employees in the lead, providing accessibility and developing AI skills at all levels.

At Turnitin, we don’t believe that using AI in the classroom is wrong or unethical. However, using AI with critical thinking and high integrity are skills that must be taught and practised. This is where AI writing detection tools play an important role providing visibility to the instructor on when AI tools may have been used.

We did not build our AI writing detection feature to be a punitive tool, nor is it intended to be used as the sole basis for any academic misconduct decision. Our AI writing detector prioritises a low false positive rate of below 1% to give educators confidence that when a document is predicted to have AI writing, they can focus on making well informed decisions on teaching, learning, feedback and academic integrity.

AI fluency and integrity in an AI centric world are two of the most important skills for students to learn today and that Turnitin’s AI writing detector is a vital tool in helping institutions educate the brightest minds of the future.

What advice can you give academics seeking to harness AI in the workplace?

AI should be designed, developed and tested by a wide and diverse range of learners and educators, not only engineers and AI scientists. It should also be continuously improved to make it more accessible, fair and beneficial to learners and educators. 

Gen AI systems are enormously powerful in helping to summarise, organise and make sense of complex thoughts and writing. I’ve found that LLMs are most useful in helping to take a set of notes or rough idea sketches to a well thought out document outline that I can then compose in my own writing style.

By helping to “cross the writing rubicon” from rough ideas to workable initial draft, LLMs are a massive time saver, allowing me to focus applying my own knowledge and judgement in the critical moments that help define a composition - whether it’s a project plan, quick email, academic writing, etc..

In general, thinking about Gen AI as an augment to our human intelligence, rather than wholesale automation and human replacement is the right way to approach bringing Gen AI into the workplace.

How do you see the role of AI evolving in the education sector, particularly within the context of academic assessment tools?

Unlike previous generations of AI such as recommender engines, Gen AI has the potential to augment human intelligence and skills, rather than just replace human labour through automation. As mechanistic knowledge work gets gradually handed off to smarter and smarter AI, skills such as collaboration, empathy and creativity will become even more sought after than today. Assessing these skills will require the ability to scale more authentic assessments - AI will play a huge role in augmenting how teachers can provide actionable and insightful feedback.

The need for more authentic assessments will require more written assessments - we believe that writing and expressing everyone’s authentic voice will be more important than ever before. At Turnitin, we’re tackling this in several ways. Our upcoming student writing tool will provide educators proof-of-process data points such as edit history, typing speed and copy-paste actions.

We’re also bringing powerful AI assistance to our Turnitin Feedback Studio grading products, enabling teachers to grade and give feedback faster and more efficiently than ever before. Looking ahead, we are excited to continue to build experiences to help students express their own brilliance and challenge themselves, both with and without AI assistance.

We think that Turnitin’s AI writing detection features and AI enhanced learning, grading and feedback tools will be invaluable assets to school districts and higher education institutions in helping to train the students of the future and we can’t wait to partner with our customers and users in building that future together.

******

Make sure you check out the latest edition of Technology Magazine and also sign up to our global conference series - Tech & AI LIVE 2024

******

Technology Magazine is a BizClik brand